impact of climate change on extreme rainfall events and ... · impact of climate change on extreme...

Impact of climate change on extreme rainfallevents and flood risk in India

P Guhathakurta∗, O P Sreejith and P A Menon

India Meteorological Department, Shivajinagar, Pune 411 005, India.∗e-mail: [email protected]

The occurrence of exceptionally heavy rainfall events and associated flash floods in many areas duringrecent years motivate us to study long-term changes in extreme rainfall over India. The analysis of thefrequency of rainy days, rain days and heavy rainfall days as well as one-day extreme rainfall and returnperiod has been carried out in this study to observe the impact of climate change on extreme rainfallevents and flood risk in India. The frequency of heavy rainfall events are decreasing in major parts ofcentral and north India while they are increasing in peninsular, east and north east India. The studytries to bring out some of the interesting findings which are very useful for hydrological planning anddisaster managements. Extreme rainfall and flood risk are increasing significantly in the country exceptsome parts of central India.

1. Introduction

In recent years, heavy precipitation events haveresulted in several damaging floods in India. Theconsecutive flash floods over three major metrocities in the same year, i.e., Mumbai in July 2005,Chennai in October 2005 and again in December2005 and Bangalore in October 2005 caused heavydamages in economy, loss of life, etc. Some of therecent studies (Goswami et al 2006; Rajeevan et al2008) on extreme rainfall events over India weremostly concentrated in central India. Regionswhere the country recently experienced extremerainfall events such as 94 cm in Mumbai, 156.5 cm(country’s highest ever recorded one-day pointrainfall) in Cherrapunji, or Chennai or Bangalorewere not included in the studies. A detailed studyon extreme rainfall events covering the entireregion and using the latest data is urgently neededto obtain a clear idea about the impact of climatechange on the extreme weather events of the coun-try. It may be mentioned that understanding the

changes in extremes weather events is more impor-tant than the changes in mean pattern for betterdisaster management and mitigation. According tothe latest report of Intergovernmental Panel on Cli-mate Change (IPCC 2007), another aspect of theprojected changes is that “wet extremes are pro-jected to become more severe in many areas wheremean precipitation is expected to increase, anddry extremes are projected to become more severein areas where mean precipitation is projected todecrease. In the Asian monsoon region and othertropical areas there will be more flooding”.

A large amount of the variability of rainfallis related to the occurrence of extreme rainfallevents and their intensities. Therefore, there is aneed to know the magnitudes of extreme rainfallevents over different parts of the area under study.The study of spatial variability of extreme rainfallevents help to identify zones of high and low valuesof extreme rainfall events. A detailed regionalizedstudy is practically useful for planners and otherusers. Sinha Ray and Srivastava (2000) have done

Keywords. Climate changes; extreme rainfall events; flood risk.

J. Earth Syst. Sci. 120, No. 3, June 2011, pp. 359–373c© Indian Academy of Sciences 359

360 P Guhathakurta et al

trend analysis of heavy rainfall events over selectedstations all over India and reported a decreasingtrend over most parts of the country. However,over northern west coast and a few stations in thenorthern parts (Haryana and Punjab), a significantincreasing trend was found. Earlier, Rakhecha andPisharoty (1996) have studied heavy rainfall eventsduring southwest monsoons for some selected sta-tions in the country. Rakhecha and Soman (1994)analysed the annual extreme rainfall series in thetimescale of 1 to 3 days at 316 stations, welldistributed over the country, covering 80 years ofrainfall data from 1901 to 1980 for trends and per-sistence using standard statistical tests. They havereported that the annual extreme rainfall recordsof most stations are free from trends and persis-tence. However, the extreme rainfall series at sta-tions over the west coast, north of 12◦N and atsome stations to the east of the Western Ghatsover the central parts of the peninsula showed asignificant increasing trend at 95% level of confi-dence. Stations over the southern peninsula andover the lower Ganga valley exhibit a decreasingtrend at the same level of significance. Stephensonet al (1999) using the data for the period Juneto September 1986–1989 have investigated extremedaily rainfall events and their impact on ensem-ble forecasts of the Indian monsoon. Most of thestudies on extreme rainfall over India (Rakhechaand Soman 1994; Sen Roy and Balling 2004) usedlimited number of stations. Sen Roy and Balling(2004) have considered parameters such as totalannual precipitation, 5-day total precipitation and30-day total precipitations in their extreme rain-fall analysis. In our study, we have considered fourindices which are directly related to extreme rain-fall events. The results are very useful for fur-ther studies and disaster management. Recently,Goswami et al (2006) using the daily gridded dataof India Meteorological Department for the period1951–2003, examined the trend of extreme rainfallover India. However their study was confined to thecentral India. In their study, they have reportedan increase in the frequency and the magnitudeof extreme rain events and a significant decreasingtrend in the frequency of moderate events over cen-tral India. It may be noted that for extreme rainfallanalysis to study the behaviour of changes in theextreme events, real or actual station data is morerealistic than the gridded dataset. In the griddeddatasets, extreme events will not be captured onmost of the occasions due to interpolation or aver-aging scheme used in gridding. This can even mis-lead the signals of hydrological extremes for bet-ter disaster management. Rajeevan et al (2008)used 104 years of gridded dataset and selecteda box mostly comprising central India similar to

Goswami et al (2006) and examined the variabilityand long-term trends of extreme rainfall events.

This study aims to analyse some of the extremerainfall indices using reliable, consistent and suf-ficient amount of rain gauge station data and tostudy the changes in the frequency of rainy days,rain days as well as heavy rainfall days. One-day extreme rainfall analysis was also performedto study the changes in the intensity of extremeweather events. Increasing flood risk is now recog-nized as the most important sectoral threat fromclimate change in most parts of the world. Tem-poral variability in extremes is analysed for fixeddecades from 1951–1960, 1961–1970, 1971–1980,1981–1990 and 1991–2000. Return period analy-sis is used to analyse the changing probability ofextreme rainfall events for observed data in orderto find out the changes in flood risk.

2. Data and methodology

Daily rainfall data for the period 1901–2005 inmore than 6000 rain gauge stations all over Indiawere considered initially. The data was collectedfrom India Meteorological Department, Pune.From a network of stations, 2599 stations having30 years or more data were selected. Selecting alarge network of stations having maximum dataavailability (Fowler et al 2000; Fowler and Kilsby2002, etc.) helps better spatial representationinstead of using fewer number of stations withuniform data availability.

Annual time series of the following indices areconstructed for these stations.

• Frequency of rainy days. According to the defi-nition followed by India Meteorological Depart-ment, a day is called ‘rainy day’, if the rainfall ofthat day is 2.5 mm or more.

• Frequency of rain days. A day is considered as‘rain day’ if the amount of rainfall of that day is0.1 mm or more.

• Frequency of heavy rainfall days (including veryheavy and extremely heavy). A day is called‘heavy rainfall day’ if the rainfall of that dayis 64.5 mm or more according to India Meteo-rological Department. This includes very heavy(i.e., 124.5–244.5 mm) and extremely heavy (i.e.,>244.5 mm) rainfall.

• Annual one-day extreme rainfall series. Annualone-day extreme rainfall is usually defined as themaximum daily rainfall within each year, so onewould have as many extreme values as the totalnumber of years.

While analysing the data it has been found thatthere were some missing observations in some years

Impact of climate change on extreme rainfall events 361

Figure 1. (a) Location of 2599 rain gauge stations over India. (b) Geographical location of India’s 36 meteorologicalsub-divisions.


Figure 2. Annual frequency of normal rainy days.

of data. For the first three cases we have removedthe effect of missing days observation by dividingthe actual annual frequency with actual number ofobservation of that year and then multiplying bythe number of days of that year. It may be men-tioned that almost the same methodology is usedby Sen Roy and Balling (2004) to remove the effectof missing years.

Recent studies indicate that the most widelyused method for detecting the trend is the non-parametric Mann–Kendall (MK) trend test. Mann(1945) originally derived the test and Kendall(1975) subsequently derived the test statistic com-monly known as the Kendall’s tau statistic. Itwas found to be an excellent tool for trend detec-tion in different applications (Hirsch et al 1982;Lettenmaier et al 1994; Burn and Hag-Elnur 2002).

Under the null hypothesis H0, a series {x1, ...,xN} are obtained from a population where the ran-dom variables are independent and identically dis-tributed (i.e., null hypothesis has no trend), theMK test statistic is:

S =N−1∑

i=1

N∑

j=i+1

sgn (xj − xi) ,

where

sgn (x) =

{+1, x > 00, x = 0

−1, x < 0.

And tau is estimated as:

τ =2S

N (N − 1).

It may be mentioned that the MK trend test is anon-parametric test which allows missing data. Itis not affected by gross data errors and outliers.However, this test can be used only as a yes/notest to detect the existence of a slope.

To determine the amount of increase or decreasein 100 years, we have adopted least square linearfit of the data. The advantage of linear regressionis that it provides an estimate of slope, confidenceinterval and quantifies goodness of fit. However, itdoes not handle missing data and may be greatlyaffected by outliers and cyclic data. As mentionedearlier in our study we have removed the effectof missing days observation by dividing the actualannual frequency with actual number of observa-tion days of that year and then multiplying by thenumber of days of that year.

To determine the flood risk, extreme values for25, 50 and 100 years return period are calculated.Probabilistic extreme value theory, which primar-ily deals with the stochastic behaviour of the maxi-mum and minimum random variables, extreme andintermediate order statistics and exceedance over(below) high (low) thresholds are determined bythe underlying distribution.

Traditionally, the three extreme value distribu-tions are applied to annual maximum daily rainfall.


(a)

(b)

Figure 3. Stations with significant increasing/decreasing trend in frequency of rain days at (a) 95% significant level and(b) 99% significant level using MK non-parametric trend test.


(a)

(b)

Figure 4. Stations with significant increasing/decreasing trend in frequency of rainy days at (a) 95% significant level and(b) 99% significant level using MK non-parametric trend test.


35º

30º

25º

20º

15º

10º

35º

30º

25º

20º

15º

10º

35º

30º

25º

20º

15º

10º

35º

30º

25º

20º

15º

10º

70º 75º 80º 85º 90º 95º

70º 75º 80º 85º 90º 95º

Increasing trend

Decreasing trendSignificant level 95%

Increasing trend

Decreasing trend

Significant level 99%

Arabian Sea

Bay of Bengal

Arabian Sea

Bay of Bengal

(a)

(b)

Figure 5. Stations with significant increasing/decreasing trend in frequency of heavy rainfall days at (a) 95% significantlevel and (b) 99% significant level using MK non-parametric trend test.


In this paper, we have used the generalized extremevalue (GEV) distribution, which has the flexibil-ity of all the other extreme value distributionsThe GEV distribution was developed by Jenkinson(1955), Hosking et al (1985) and Galambos (1987).The cumulative distribution function (cdf) of theGEV distribution is as follows.

F (x; ψ, β) = exp(−

(1 + ζ

x − ψ

β

)−1/ζ),

for 1 + ξ(x − ψ)/β > 0, where ψ, β and ξ arereferred to as the location, scale and shape para-meters, respectively. The particular case of (1) forξ → 0 is the Gumbel distribution

F (x; ψ, β) = e−e(ψ−x)/β

.

The median is ψ − β ln (− ln (1/2)).The mean is ψ + γβ where γ=Euler–Masche-

roni constant ≈0.5772156649015328606.

(a) Rain days

36ºN

33ºN

30ºN

27ºN

24ºN

21ºN

18ºN

15ºN

12ºN

9ºN

6ºN

66ºE 69ºE 72ºE 75ºE 78ºE 81ºE 84ºE 87ºE 90ºE 93ºE 96ºE 99ºE

100

80

60

40

20

0

− 20

−40

−60

−80

−100

−120

(b) Rainy days

36ºN

33ºN

30ºN

27ºN

24ºN

21ºN

18ºN

15ºN

12ºN

9ºN


80

60

40

20

0

−20

−40

−60

−80

−100

−120

6ºN

Figure 6. Increase/decrease in frequency of (a) rain days (b) rainy days and (c) heavy rainfall days in 100 years.


(c) Heavy rainfall days

36ºN

33ºN

30ºN

27ºN

24ºN

21ºN

18ºN

15ºN

12ºN

9ºN

6ºN


70

50

40

20

10

5

0

−5

−10

−20

Figure 6. (Continued).

The standard deviation is:

βπ/√

6.

The mode is ψ.Then from the Gumbel distribution with para-

meters ψ and β, the extreme value (return value)x for a period of n years can be derived by thefollowing formula:

x = ψ − β ln [− ln (F )] ,

where ψ = average − γβ (where γ is Euler’s con-stant, approximately 0.557).

β = 0.78σ (where σ is the standard deviation),F is the random variate drawn from the uniformdistribution in the interval [0, 1] = (n − 1)/n.

3. Result and analysis

India lies within the tropical monsoon zone andhas a high spatial variation in rainfall and its fre-quencies. Figure 1(a) shows the location of 2599rain gauge stations across the country which wereconsidered for this study, while figure 1(b) showsthe 36 meteorological sub-divisions of India. Fromthe annual frequency of rainy days, rain days andheavy rainfall days; average or normal rainy days,rain days and heavy rainfall days were computed.High spatial variations in the climatology of fre-quency of rainy days can be seen in figure 2.Annual normal rainy days vary from the low valueof 10 over extreme western parts of Rajasthan to

the high frequency of 130 days over northeasternparts of the country. The northeastern parts of thecountry as well as sub-Himalayan West Bengaland also extreme western coast line of the countryreceived an average of more than 100 rainy daysin a year.

Significant decreasing trends in the frequency ofrain days, rainy days and heavy rainfall days canbe seen over most parts of central, north and east-ern parts of the country using both MK and t tests.Figure 3(a and b) shows the decreasing/increasingtrend of stations having significant at 95% and99%, respectively in the frequency of rain daysobtained by using the MK test. Wet days haveincreased in peninsular India particularly overKarnataka, Andhra Pradesh and parts ofRajasthan and some parts of eastern India, whilemost parts of central and northern India showed adecreasing trend in frequency of rain days. Figure4(a and b) shows the decreasing/increasing trendof stations having significant at 95% and 99%,respectively in the frequency of rainy days obtainedby using the MK test. Figure 5(a and b) showsthe decreasing/increasing trend of stations havingsignificant at 95% and 99%, respectively in the fre-quency of heavy rainfall days obtained by using theMK test. Almost similar patterns of trends thatwere observed in frequency of rain days in figure 3are also observed in the frequency of rainy days infigure 4 with some differences. Significant increas-ing trends are clearly noticed in the frequency ofrain days over eastern Rajasthan, however, these


increasing trends were not maintained in the fre-quency of rainy days. These imply that althoughnumber of wet days are increasing over easternRajasthan, occurrences of rain are not intensified.Increasing trends in the frequency of heavy rain-fall days are observed in very few stations mostlyin Konkan and Goa and adjoining regions of west-ern coast and some isolated stations of easternIndia. Frequency of heavy rainfall events decreasedsignificantly for most parts of the country.Figure 6 shows the contour analysis on the rateof decrease/increase, i.e., slopes in the frequencyof rain days, rainy days and heavy rainfall daysin 100 years significant at 95% level obtained byusing linear regression. The rate of increase in raindays has been observed to be around 40 to 50 daysin 100 years in peninsular India particularly overKarnataka and Andhra Pradesh. Increase in raindays has also been observed over most parts ofRajasthan, parts of gangetic West Bengal andadjoining parts of Jharkhand. This indicates thatthe great desert areas of India are becoming wet.It may be mentioned that Guhathakurta andRajeevan (2008) have also reported an increasingtrend in the annual mean rainfall of these regions.However, extreme southern parts of the country,

i.e., Kerala and Tamil Nadu are experiencing moredry days. Decreasing trends in annual mean rain-fall over Kerala and Tamil Nadu have also beenreported by Guhathakurta and Rajeevan (2008).Figure 6(c) clearly indicates that frequency ofextreme rainfall events has decreased in most ofcentral and north India. It may be mentioned thatGoswami et al (2006) reported significant increas-ing trends in the frequency of extreme events overcentral India using gridded data. This is becausethey have studied extreme rainfall with a griddeddataset. Maximum increase in the frequency ofheavy rainfall events (20 to 50 days in 100 years)has been noticed in northeast India and coastalKarnataka. The country’s highest observed one-day point rainfall and also world’s highest two-daypoint rainfall occurred in Cherrapunji of northeastIndia in 1995 (Guhathakurta 2007).

Figure 7 gives the spatial pattern of India’s high-est one-day ever recorded point rainfall. Occur-rences of 40 cm or more rainfall has been noticedalong the west and east coast of India, GangeticWest Bengal and northeastern parts of India.Trend analysis was performed over an annual one-day extreme rainfall series for each of the stations.Figure 8(a and b) shows the increasing/decreasing

70º 75º 80º 85º 90º 95º

30º

35º

25º

20º

15º

10º

5º70º 75º 80º 85º 90º 95º

5º

10º

15º

20º

25º

30º

35ºOne-day maximum rainfall

0 10 20 30 40 50 60 70 80 90 100

Figure 7. Highest one-day extreme rainfall in India.


35º

30º

25º

20º

15º

10º

35º

30º

25º

20º

15º

10º

35º

30º

25º

20º

15º

10º

35º

30º

25º

20º

15º

10º

70º 75º 80º 85º 90º 95º

70º 75º 80º 85º 90º 95º

Increasing trendDecreasing trend


Increasing trendDecreasing trend


Arabian Sea

Bay of Bengal

Arabian Sea

Bay of Bengal

(a)

(b)

Figure 8. Stations with significant increasing/decreasing trend in one-day extreme rainfall at (a) 95% significant level and(b) 99% significant level using MK non-parametric trend test.

trend of stations having significant at 95% and99%, respectively in the one-day extreme rain-fall obtained by using the MK test. The non-

parametric test shows significant increasing trendin one-day extreme rainfall over the south penin-sular region, Maharashtra, Gujarat region, Bihar


36ºN

33ºN

30ºN

27ºN

24ºN

21ºN

18ºN

15ºN

12ºN

9ºN

6ºN66ºE 69ºE 72ºE 75ºE 78ºE 81ºE 84ºE 87ºE 90ºE 93ºE 96ºE 99ºE

20

15

10

5

0

−5

Figure 9. Increase/decrease in one-day extreme rainfall (cm) in 100 years.

36oN

33oN

30oN

27oN

24oN

21oN

18oN

15oN

12oN

9oN

6oN

66oE 69oE 72oE 75oE 78oE 81oE 84oE 87oE 90ºE 93oE 96oE 99oE

30

25

20

15

10

7

Figure 10. Extreme values for 25-year return period.


1951−1960

1961−1970

1971−1980

1981−1990

1991−2000

36ºN

33ºN

30ºN

27ºN

24ºN

21ºN

18ºN

15ºN

12ºN

9ºN


30

25

20

15

10

7

36ºN

33ºN

30ºN

27ºN

24ºN

21ºN

18ºN

15ºN

12ºN

9ºN

6ºN


30

25

20

15

10

7

36ºN

33ºN

30ºN

27ºN

24ºN

21ºN

18ºN

15ºN

12ºN

9ºN


30

25

20

15

10

7

36ºN

33ºN

30ºN

27ºN

24ºN

21ºN

18ºN

15ºN

12ºN

9ºN


30

25

20

15

10

7

36ºN

33ºN

30ºN

27ºN

24ºN

21ºN

18ºN

15ºN

12ºN

9ºN


30

25

20

15

10

7

Figure 11. Decadal variability of extreme values (cm) for 25-year return period.


0.0

5.0

10.0

15.0

20.0

25.0

30.0

35.0

7.0−10.0 10.0−15.0 15.0−20.0 20.0−25.0 25.0−30.0 >30.0

1951−19601961−19701971−19801981−19901991−2000

Figure 12. Percentage frequencies of extreme values (cm) for 25-year return period in different range for each of the decades.

and some other isolated areas. The parametric testusing linear regression helps to find out the rateof increase in 100 years and these are shown infigure 9. Significant decrease both in intensity andfrequency of extreme rainfall have been observedover Chattisgarh, Jharkhand and some parts ofnorth India.

The extreme values for 25 return periods showedlarge variation (figure 10). Values of a 25-yearreturn period varied from 10 cm over Rajasthan,parts of north interior Karnataka and Marathwadato 30 cm over West Bengal, Orissa and parts ofnorth east India. To find the change in flood risk,we have studied the decadal variability of a 25-yearreturn period. Temporal variability in extremes hasalso been analysed in fixed decades from 1961–1970, 1971–1980, 1981–1990 and 1991–2000 earlierby Fowler and Kilsby (2003) for investigating theflood risk in the United Kingdom although largeuncertainty prevails along with change of time. Fig-ure 11 shows the decadal variability of extremevalues (cm) for a 25-year return period for thedecades 1951–1960, 1961–1970, 1971–1980, 1981–1990 and 1991–2000. In the decade 1991–2000,the probability of occurrence of extreme values ofmore than 25 cm or more was found along thenorthern parts of the west coast, throughout theeastern coast, West Bengal and Gujarat regions.Figure 12 brings out the percentage frequencies ofextreme values (cm) for a 25-year return periodin different ranges for each of the decades. Per-centage frequencies for different ranges are calcu-lated from the decadal return values of all the2249 rain gauge stations. It is clear that fre-quency of extreme values greater than 30 cm weremore in the recent two decades, viz., 1981–1990

and 1991–2000. The evidence of increase inflood risk during the recent decades is clearlynoticed.

4. Conclusions

This study reveals the noticeable changes in theextreme rainfall events that occurred over India inthe past century. The country experienced largespatial variations in annual normal rainy days.Annual normal rainy days varied from 10 days overextreme western parts of Rajasthan to the highfrequency of 130 days over northeastern parts ofthe country. The non-parametric test as well as thelinear trend analysis identified decreasing trendsin the frequency of wet days in most parts of thecountry. Trend analysis of frequency of rain days,rainy days and heavy rainfall days showed signifi-cant decreasing trends over central and many partsof north India; and increasing trends over penin-sular India. Also, the great desert areas of thecountry have experienced increased number of wetdays. Analysis of one-day extreme rainfall serieshas shown that the intensity of extreme rainfallhas increased over coastal Andhra Pradesh andits adjoining areas, Saurashtra and Kutch, Orissa,West Bengal, parts of northeast India, and eastRajasthan. Significant decrease in intensity as wellas frequency of extreme rainfall have been observedover Chattisgarh, Jharkhand and some parts ofnorth India. The flood risk also increased signifi-cantly over India. The flood risk was more in thedecades 1981–1990, 1971–1980 and 1991–2000. Theincrease of flood risk has increased during the lasttwo decades mostly over the eastern coast, West


Bengal, east Uttar Pradesh, Gujarat and Konkanregion.

Acknowledgements

The authors thank the Director General of Mete-orology, India Meteorological Department for hisencouragement and support in completing thiswork. They also thank the referees for their fruitfulsuggestions and comments to improve the qualityof the paper.

References

Burn D H and Hag-Elnur M A 2002 Detection of hydrologictrend and variability; J. Hydrol. 255 107–122.

Fowler H J and Kilsby C G 2002 Precipitation and theNorth Atlantic Oscillation: A study of climatic variabilityin Northern England; Int. J. Climatol. 22 843–866.

Fowler H J and Kilsby C G 2003 A regional frequency analy-sis of United Kingdom extreme rainfall from 1961 to 2000;Int. J. Climatol. 23 1313–1334.

Fowler H J, Kilsby C G and O’Connell P E 2000 Astochastic rainfall model for the assessment of regionalwater resource systems under changed climatic condi-tions; Hydrol. Earth Syst. Sci. 4 263–282.

Galambos J 1987 The Asymptotic Theory of Extreme OrderStatistics, 2nd edn. (Melbourne, Florida: Krieger).

Goswami B N, Venugopal V, Sengupta D, MadhusoodananM S and Xavier P K 2006 Increasing trend of extremerain events over India in a warming environment; Science314 1442–1445.

Guhathakurta P 2007 Highest recorded point rainfall overIndia; Weather 62 12.

Guhathakurta P and Rajeevan M 2008 Trends in rainfallpattern over India; Int. J. Climatol. 28 1453–1469.

Hirsch R M, Slack J R and Smith R A 1982 Techniquesof trend analysis for monthly water quality data; WaterResour. Res. 18 107–121.

Hosking J R M, Wallis J R and Wood E F 1985 Estimation ofthe generalized extreme value distribution by the methodof probability weighted moments; Technometrics 27251–261.

IPCC 2007 Fourth Assessment Report (AR4) of the UnitedNations Intergovernmental Panel on Climate Change.

Jenkinson A F 1955 The frequency distribution of the annualmaxima (or minima) values of meteorological elements;Quart. J. Roy. Meteor. 81 158–171.

Kendall M G 1975 Rank Correlation Methods (London:Griffin).

Lettenmaier D P, Wood E F and Wallis J R 1994 Hydro-climatological trends in the continental United States,1948–1988; J. Climate 7 586–607.

Mann H B 1945 Nonparametric tests against trend; Econo-metrica 13 245–259.

Rajeevan M, Bhate J and Jaswal A K 2008 Analysis of vari-ability and trends of extreme rainfall events over Indiausing 104 years of gridded daily rainfall data; Geophys.Res. Lett. 35 L18707, doi: 10.1029/2008GL035143.

Rakhecha P R and Pisharoty P R 1996 Heavy rainfall duringmonsoon season: Point and spatial distribution; Curr. Sci.71 177–186.

Rakhecha P R and Soman M K 1994 Trends in the annualextreme rainfall events of 1 to 3 days duration over India;Theor. Appl. Climatol. 48(4) 227–237.

Sen Roy S and Balling R C 2004 Trends in extremedaily precipitation indices in India; Int. J. Climatol. 24457–466.

Sinha Ray K C and Srivastava A K 2000 Is there any changein extreme events like drought and heavy rainfall?; Curr.Sci. 79(2) 155–158.

Stephenson D B, Rupa Kumar K, Doblas-Reyes F J,Royer J F, Chauvin F and Pezzulli S 1999 Extremedaily rainfall events and their impact on ensemble fore-casts of the Indian monsoon; Mon. Weather Rev. 1271954–1966.

MS received 22 July 2010; revised 20 January 2011; accepted 21 January 2011

impact of climate change on extreme rainfall events and ... · impact of climate change on extreme...

Documents